Brevitas: neural network quantization in PyTorch
-
Updated
Aug 28, 2026 - Python
Brevitas: neural network quantization in PyTorch
Low Precision(quantized) Yolov5
Low-Precision Neural Networks for Classification on PYNQ with FINN
Offline pipeline for the 91-parameter quantised CNN which replaces Optimal Filtering in the ATLAS Tile Calorimeter readout at HL-LHC pile-up. Preprocessing, knowledge distillation, quantisation-aware training and firmware export via Brevitas, QONNX and hls4ml. GSoC 2026, CERN-HSF.
1-bit binarized CNN (CNV-W1A1) on PYNQ-Z1 via Brevitas + FINN: 1367 fps, 0 DSP, 531 fps/W measured on hardware
(1/3) Training and quantization of a neural network model in the master thesis.
Portable agent skill to install Brevitas (LLM metering/optimization gateway) from Claude Code, OpenAI agents, or any tool. Run /install-brevitas.
Add a description, image, and links to the brevitas topic page so that developers can more easily learn about it.
To associate your repository with the brevitas topic, visit your repo's landing page and select "manage topics."